Tip for Cursor IDE users: use Agents. md as your single source of truth for AI agent behavior. Don't just rely on rules. think of it as a README written for agents rather than humans. It tells coding agent how to build, test, lint, structure work, and respect conventions across your codebase. So you don't have to spoon feed everything. It reduces rules file chaos across different tools whether you are using claude code or cursor. What Cursor does with agents. md? → When cursor’s agent runs on a project: It searches for agents. md in the repository (and often the nearest one in nested folders in monorepos). → It uses the content to: - understand how to install dependencies, run dev/build/test commands. - follow code style, folder structure, and architectural constraints. What to put in agents. md? → project overview Tech stack, entry points, critical scripts, env notes → build and run Commands to install, build, and run locally/CI → testing How to run tests, required flags, coverage thresholds, known flaky suites → lint/format Tools, commands, autofix flags, style rules that matter → code generation don’t edit zones, how to regenerate clients/types/migrations → architecture and conventions Module boundaries, naming, folder layout, patterns to prefer/avoid → security and secrets What not to log/edit, redaction rules, .env handling, secret scanning → performance and quality gates Benchmarks to run, budgets, perf-sensitive paths → release process Versioning, changelogs, PR checks, release scripts Checkout this example snippet: